Unfortunately i haven't had time to test openmpi 2.0.1 very much, but it will be the version of choice here (regardless of what foss-2017a comes up with). I basically got slowed down by the fact that 80% of the software requests we got when opening up the new system has been stuff that wasn't in EB before. So i've had to create stuff and then i went on vacation, which ends after New Year.
I will try to move forward with the SW requests we have so i can start looking into things like problems with OpenMPI 2.x. But it's going to be somewhat slow... On 12/09/2016 06:55 PM, Kenneth Hoste wrote: > Hi Åke, > > On 07/12/16 20:03, Åke Sandgren wrote: >> Since i couldn't join todays meeting, here are some comments on the >> outcome. >> >> Regarding the suggested foss/2017a, OpenMPI 2.0.1 is way better than >> 1.10 in my opinion. Esp since i've seen cases of 1.10.3 failing to >> actually start... probably a weird corner case, but it's repeatable. >> >> FFTW 3.3.5 (at least for this toolchain) should be compiled with >> --enable-sse2 --enable-avx --enable-avx2 >> and preferably with --enable-avx512 too. >> >> 3.3.5 built that way is safe to run on anything from sse2-only and >> upwards. >> And 6.2(3) is supposed to be able to build it that way even on sse2-only >> systems. > > Thanks for the feedback! > > Have you seen any big problems pop up with using OpenMPI 2.0.1, due the > backwards-incompatible changes with OpenMPI 1.10.x? > What have you tried so far on top of a foss toolchain using OpenMPI 2.0.1? > > The suggestions w.r.t. FFTW make a lot of sense, thanks for the reminder. > > I plan to open the pull requests for the 2017a toolchain soon, I will > definitely leave ample time to provide feedback there too. -- Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden Internet: [email protected] Phone: +46 90 7866134 Fax: +46 90-580 14 Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se
